I have been trying out ChatGPT to explore its capabilities. And my one-sentence conclusion after using it to ask some very basic/simple greenhouse gas emission-related questions: Use ChatGPT with extreme caution and expect errors, even on simple data questions.
Here is a slightly expanded explanation. I asked ChatGPT about the GHG emissions for China, the US, and the globe . According to ChatGPT, China’s share of global GHG emissions was 28%, while the US share was 15% for 2021. In terms of absolute number, it also indicated that China’s 2019 emissions were 10.17 gigatons of carbon dioxide carbonequivalents and US emissions in 2019 were 5.41 gigatons (based on information from the Global Carbon Project). This seemed plausible (and was close to what I found when checking the Global Carbon Project website directly). However, it then gave global GHG emissions for 2019 at 59.1 gigatons – that of course does not plausibly match up with the numbers and percentages for the US and China. Based on the 2019 US and China numbers, global emission should be to 36 gigatons.
It seems likely that it served up numbers that encompassed different GHG sources, i.e. possibly including land use emissions in one set of numbers and ignoring them in the other. In any event, I was surprised that Chatgpt provided internally inconsistent answers. When challenged, here is what ChatGPT said:
“You are correct that the numbers I provided in my previous answers were inconsistent. I apologize for the confusion and any misinformation that may have caused. According to the latest available data from the Global Carbon Project, China’s greenhouse gas (GHG) emissions in 2019 were approximately 14.1 gigatons of carbon dioxide equivalent (GtCO2e), which is about 30% of global emissions. The United States was the second-largest emitter with approximately 4.9 GtCO2e or about 11% of global emissions. Global GHG emissions in 2019 were indeed around 59 GtCO2e, as you mentioned. The largest sources of global GHG emissions are energy-related emissions from burning fossil fuels for electricity, transportation, and industry, as well as emissions from agriculture, land use, and waste management. Again, I apologize for any confusion my previous answers may have caused, and I hope this clarifies the correct information about global and Chinese GHG emissions in 2019.”
Better explanation, but … still 🤔 ｢(ﾟﾍﾟ) ｢(･ิ.･ิ) ｢(‘◉⌓◉’)
Of course, I should have noted that many people have reported ChatGPT making serious mistakes and providing wildly wrong answers. But I thought a simple data information request should be a softball – not much different from a traditional search engine query. But clearly not so. Too bad that ChatGPT cannot even be trusted for some simple data queries.